A new threat intelligence report by DuckDuckGoose indicates that generative AI is enabling the creation of synthetic identities at a pace exceeding the adaptation capabilities of current verification infrastructure across Latin America’s digital banking ecosystems. The report, titled “When Identity Becomes Generatable,” analyzes how generative AI has shifted identity creation from manipulation to real-time synthetic human generation, capable of bypassing existing onboarding and authentication systems. This development poses a structural risk to digital banking.
The report highlights the rapid proliferation of generative AI tools, noting that in Q4 2025 alone, more than 55 synthetic media generators were released. Since early 2024, image-to-video generation technology has expanded by over 1,000%, which erodes the underlying assumptions of biometric and liveness verification systems. Monitoring efforts identified nearly 868,000 fine-tuned synthetic variants being created monthly across open AI ecosystems, many of which are specifically optimized to bypass onboarding and authentication controls.
Each new variant introduces unfamiliar identity characteristics, leading to exposure windows where verification systems encounter synthetic identities they have not previously detected. This acceleration is particularly evident in Brazil and across Latin America. In these regions, factors such as PIX-driven real-time payments, large-scale remote onboarding processes, and the rapid expansion of neobanks have made reliable identity verification a critical financial control. Brazilian institutions reported R$10.1 billion in banking fraud losses in 2024, with fraud execution timelines increasingly compressing to minutes rather than days.
The report concludes that the primary threat is now synthetic presence—AI-generated humans interacting with verification systems in real time—rather than merely manipulated media. While traditional Know Your Customer (KYC) and liveness controls function as designed, they were not architected to address continuously generated identities that adapt faster than existing detection models can evolve. As the velocity of generator releases increases, verification environments face a persistent detection gap, signaling a shift from periodic security upgrades to a demand for continuously adaptive infrastructure for identity assurance.
“Identity verification was built on the assumption that visual presence implied a real person,” stated Parya Lotfi, CEO of DuckDuckGoose. “Generative AI removes that assumption. Identity has become generatable. Trust must be established at the moment of identity creation, not after fraud occurs. That transition is already underway across Latin America’s digital financial systems.”
For identity platforms operating in digital economies such as Brazil, the report suggests that synthetic media detection is transitioning from an advanced capability to a necessary infrastructure control, alongside established methods like document verification and liveness checks.
DuckDuckGoose specializes in explainable deepfake detection technology. The company assists identity platforms and financial institutions in verifying biometric authenticity within large-scale onboarding and authentication environments.
